RFR: 8294270: make test passes awkward -status:-status:error, fail to jtreg

[Aleksey Shipilev]

[JDK-8230067](https://bugs.openjdk.org/browse/JDK-8230067) added jtreg retries, but it does so awkwardly: it sets `export JTREG_STATUS="-status:error,fail";` on retry, and then uses it as `-status:$$$${JTREG_STATUS}`.

Which means we end up passing either: `-status:` or `-status:-status:error,fail` to jtreg. I confirmed this by also instrumenting the local jtreg build.

Now, it is not a problem for current jtreg, because it checks the status codes awkwardly with `String.contains`:
https://github.com/openjdk/jtreg/blob/aeb552e6df73e039e20de59b3ec847f36ab6e202/src/share/classes/com/sun/javatest/regtest/tool/Tool.java#L1635-L1640

But it will be a problem if jtreg ever starts to validate the arguments properly. I propose we fix this in current build system.

Additional testing:
 - [x] Eyeballing test reports from artificially intermittently failing tests

-------------

Commit messages:
 - Fix

Changes: https://git.openjdk.org/jdk/pull/10404/files
 Webrev: https://webrevs.openjdk.org/?repo=jdk&pr=10404&range=00
  Issue: https://bugs.openjdk.org/browse/JDK-8294270
  Stats: 1 line in 1 file changed: 0 ins; 0 del; 1 mod
  Patch: https://git.openjdk.org/jdk/pull/10404.diff
  Fetch: git fetch https://git.openjdk.org/jdk pull/10404/head:pull/10404

PR: https://git.openjdk.org/jdk/pull/10404